The alignment process using DTW for measurement of ability to reproduce perceived rhythm. The resulting alignment path may be visualized as a low valley of Euclidean distance scores between speech spectrum segments of patients and speech therapist template recording, meandering through the hilly landscape of the matrix, beginning at (0,0) and ending at the final point (X,Y).
The top part of the figure shows a selected example of probability densities. The bottom of the figure shows the result of Wald task classification. As a result, the dark-gray shaded bars are the regions in which the feature x assumes the value from the subset X1 and the light shaded bars are the areas predicted for the subset X2. The black bars represent the indecisive situation, for which observation x ∈ X0 does not provide sufficient information for a safe decision regarding one of the subsets X1 or X2.
The vowel triangle space areas for subjects with PD (up) and HC subjects (down).
Details of new articulation measures performed on rapid steady syllable repetition. (a and b) Speech signals of rapid steady /pa/-/ta/-/ka/ syllables repetition; (c and d) light gray lines represent obtained intensity sequences, dashdot lines represent the RRIS. The RIRV is computed as SD from the obtained intensity; (e and f) SDCV; (g and h) RFPC. The left panel is for a person with PD, the right panel is for a HC subject.
Selected probability densities of all representative measures (features) in preparation for Wald task classification. The vertical axes are the probability densities P(“measure”) of the normalized features values of each measure. The dashdot lines are for HC speakers, the solid lines for PD subjects.
Details of affected speech dimensions in PD: PH, phonation; AR, articulation; PR, prosody.
List of participants with sex, age, and duration of disease prior to recording. Entries labeled “n/a” are for HC, for which duration of disease is not applicable.
List of the vocal tasks.
Overview of measurement methods used as features applied to acoustic signals recorded from each subject.
List of results of all measures with mean values, SD values, correlations between the measurements methods, statistical significances, and summaries of retained measures. See main text for detailed description of the algorithm used to calculate these results.
List of Wald task classification values for all pre-selected measures and values of speech therapist evaluation. See main text for detailed description of the algorithm used to calculate these results. The total sum of the overall performance value represents the speech performance of each subject: More positive points are associated with higher progression of PD vocal impairment; the negative points show healthy speech performance. The total sum of ratings represents the suitability of the measure in differentiating PD patients from HC. The speech dimension was designated as affected when it reached at least two points of voice impairment assessment. The speech therapist evaluation represents rating scores of full vocal assessment: 1 point = intact speech performance, 7 points = progressed speech impairment.
Article metrics loading...
Full text loading...